biondizzle

0 Followers · 0 Following

• Joined on 2025-12-10

Repositories

25

Projects Packages Public Activity Starred Repositories

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:38:26 +00:00

059c2e6cd9 D1: P store as BF16 using PV A-fragment layout

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:36:42 +00:00

2efd6be8af D1: P store uses tOrP0.layout (PV A-fragment TMEM layout)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:35:51 +00:00

7751eab711 D1 fix: P store uses PV A-fragment layout (p_tmem_s.outer)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:34:00 +00:00

fe1826b0de D1: test raw unnormalized output via epilogue_tma_store

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:32:55 +00:00

091cb59be5 test: paired atoms epilog from old commit 6ee28d8

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:29:52 +00:00

f23d55fd3f D1: paired atoms epilogue (no TMEM round-trip)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:26:17 +00:00

7df3c7c952 d1: sweep hd=64,128,256

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:25:31 +00:00

81378133cc fix: use mV.iterator

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:25:00 +00:00

a66a9efd4c fix: use mQ not q for LayoutEnum

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:24:17 +00:00

d2aaab5a32 d1: add diagnostic script

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:22:26 +00:00

a2d063a48b D1: N-tile support for HEAD_DIM>256

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:20:49 +00:00

7bc097163d d1: add hd=512 test

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:20:13 +00:00

32995c2ba3 d1: add quick regression test (hd=64 only)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:19:56 +00:00

eed981bee5 D1: Parameterize HEAD_DIM in FmhaKernel (64→512)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:10:43 +00:00

1a6c5e3822 docs: revised Stage D/E plan — indexer removes paged TMA, one kernel for CSA/HCA/SWA, sink merge

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:05:10 +00:00

a846193c4a cleanup: remove archive/ (240 stale files), stale example9/10, fix test table, add Stage D plan

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 03:01:04 +00:00

f3d0d67ae9 docs: update README with Stage C TMEM layout mismatch findings and status

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 02:54:55 +00:00

9c331de7ba fix: revert to composition layout for hand-constructed atoms (matching CUTLASS)

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 02:54:02 +00:00

3a2d3c66da fix: use logical_divide (not composition) for O rescale/normalize atoms to match get_tmem_load_op layout

biondizzle pushed to master at biondizzle/nvfp4-megamoe-kernel 2026-05-23 02:50:55 +00:00

3aba5cc6da fix: add NO-OP TMEM round-trip to re-map O from MMA to epilog layout